2002 CTOS Annual Meeting Oral Presentations — Biology

MICROARRAY ANALYSIS OF MALIGNANT FIBROUS HISTIOCYTOMA
[Abstract ID: 73]

Category: Biology

Presentation: Oral

Authors: Jay S Wunder¹, Nalan Gokgoz², Sasha Eskandarian², Wenqing He², Shelley B Bull², Robert E Turcotte⁵, Vivien H Bramwell⁴, Robert S Bell¹, Rita Kandel³, Irene L Andrulis²

Author Institutions: ¹University Musculoskeletal Oncology Unit Mount Sinai Hospital Toronto, Ontario, Canada; ²Samuel Lunenfeld Research Institute Mount Sinai Hospital Toronto, Ontario, Canada; ³Department of Pathology Mount Sinai Hospital Toronto, Ontario, Canada; ⁴London Regional Cancer Centre London, Ontario, Canada; ⁵University of Montreal Montreal, Quebec, Canada

Presenter: Jay S Wunder
wunder@mshri.on.ca

Correspondent: Jay S Wunder
wunder@mshri.on.ca
Toronto Ontario Canada M5G 1X5
Ph: 416-586-8807
Fax: 416-586-8397

Objectives: Malignant fibrous histiocytoma (MFH) is the most common type of soft tissue sarcoma but is poorly understood. There are few accurate predictors of outcome to guide treatment decisions. We used microarray analysis of gene expression to identify prognostic markers for MFH.

Methods: 40 MFH tumor specimens with clinical data were chosen from a prospective tumor bank from patients who did not receive preoperative chemotherapy or radiotherapy. Frozen specimens were assessed histologically to confirm viable tumor. Tumor and control RNA were indirectly labelled with fluorescent tags and simultaneously hybridized to 19K microarray slides. Arrays were scanned, quantitated and normalized prior to statistical analysis using GenePix and SPlus, and then analysed with BrB ArrayTools.

Results: We selected 111 “candidate” genes from a variety of biological pathways with potential importance in MFH, and examined their ability to discriminate between clinicopathologic variables, stage, and oncologic outcome. For each clinical variable, a group of 5-13 genes were identified which distinguished between categorical strata (F-test p<0.05). For example, 10 genes differentiated between patients who did or did not develop metastases, including TOK1 (p21-binding protein), SEI1 (cdk4-binding protein), CA1A (collagen-related gene), IRF1 (interferon regulatory factor1) and MMP9. However, rigorous assessment of prediction error using cross-validation techniques suggested that combinations of the above genes did not significantly improve prediction, recognizing the low power and small size of this sample.

Conclusions: This study suggests that the candidate gene list approach does not provide the most accurate method for class prediction for MFH. We are presently undertaking tumor comparison using an expanded microarray 19K gene set without predetermined gene selection which will likely be a more powerful approach to identify prognostically important genes in MFH.